Centre for Market and Public
Organisation
Poorer children’s educational attainment:
how important are attitudes and behaviours?
Report for the Joseph Rowntree Foundation
Edited by
Alissa Goodman and Paul Gregg
Contributors from IFS, CMPO and University of Oxford
Background and Motivation
• Children growing up in poor families end up with lower
educational attainment than children growing up in rich families
• Strong contributor to patterns of social mobility
– Low income  poor attainment  low income
• Gaps start very early in life, but tend to widen throughout school
© Institute for Fiscal Studies
What we do
• Chart socio-economic gradient in attainment across childhood
• Investigate contribution of parent and child behaviours, attitudes
to education and aspirations to the evolution of this gradient:
– Early years: home learning environments, parenting styles, healthrelated behaviours
– Primary school: lasting influence of early years, maternal aspirations,
child’s own ability beliefs
– Teenage years: young person’s own attitudes and behaviours; lasting
influence of parents; material resources in the home
– Intergenerational factors: parents’ and grandparents’ attitudes;
transmission of ability
• Assess implications for policy
Summary of data sources, and test scores used
for analysis (1)
Millenium Cohort Study
(MCS)
B
A
S
3
5
Longitudinal Study of
Young People in England
(LSYPE)
Avon Longitudinal Study
of Parents and Children
(ALSPAC)
B
A
S
4
English cohort,
born 1989/90,
sample 13,500
Avon cohort,
born 1991/92,
sample 7,800
UK cohort,
born 2000/01,
sample
11,000
K
S
K
2
S
2
K
S
1
6
7
8
9
10
Age
11
K
S
4
K
S
3
12
13
14
15
16
Summary of data sources, and test scores used
for analysis (2)
Everyone born in
Great Britain in one
week in April 1970
interviewed
Age 34 – Half of those
with children randomly
selected for parent-child
interviews
British Cohort Study
(BCS)
Children of BCS
Age Normalised
BAS Scores
3
4
5
6
7
8
9
10
Age
11
12
13
14
15
16
Measuring socio-economic position
• Aim is to capture the longer-term material resources of the household
– Log equivalised household income (averaged across points in time)
– Reported experience of financial difficulties
– Mother’s and father’s occupational class
– Housing tenure
• The measure is constructed using principal-components analysis
• Individuals are then placed into quintiles (fifths) of the population
ranked by this measure.
Educational gaps across childhood
Percentile of the test score distribution
80
70
60
50
40
30
20
Age 3 (MCS)
Highest
Age 5 (MCS)
Age 7 (ALSPAC)
Quintile 4
Age 11
Age 14 (LSYPE) Age 16 (LSYPE)
(ALSPAC/LSYPE)
Quintile 3
Quintile 2
Lowest
Decomposing these gaps: framework for analysis
• Starting point is relationship between SEP and attainment at each age
• Decompose the gap between rich and poor students into the ‘direct
effects’ of:
•
–
Family background: parental education, family demographics
–
Aspirations, attitudes and behaviours: varying at each age
Factors will explain a larger proportion of the gap if:
–
Factors is highly correlated with socio-economic position
–
Factor has a large effect upon outcomes conditional on all observables
•
Development from previous age assessed through inclusion of prior
attainment
•
Important note: this study highlights statistical associations, and does
not imply causation.
Preview of findings
• The gaps between rich and poor children is already large at age 3
continues to widen until age 14
• The following factors seem to have an important role in explaining the
perpetuation of these gaps:
–
–
–
–
–
Early home learning environment
Expectations/ aspirations for education
Beliefs in own actions making a difference
Behaviour
Material factors
• Suggests a potentially important role for policy if it can be shown that:
– More positive attitudes and behaviours cause higher attainment
AND
– Attitudes and behaviours can be influenced
From birth to age 5
Lorraine Dearden, Luke Sibieta (IFS) and Kathy Sylva (University of Oxford)
Explaining the socio-economic gradient in the
early years
• Define set of family background and possible transmission
mechanisms (“early childhood caring environment”)
• Family background
– Socioeconomic position (SEP)
– Parental education
– Demographic, and other family background
• Early childhood caring environment
– Family Interactions (mother-chid and between parents)
– Health and Well-being (birth-weight, gestation, post-natal depression)
– Childcare usage
– Home-learning environment (reading, ABCs, numbers, nursery rhymes)
– Parenting Style/Rules (bed-times, meal-times)
© Institute for Fiscal Studies
32.2
35
30
2
Number of siblings
Mother's age at birth
Selected differences in characteristics at age 3 & 5
25.0
25
20
15
10
5
0
1.6
1.5
1.1
1
0.5
0
Mother's age at birth
Number of siblings at age 5
100%
92%
79%
Per cent
80%
98%
85%
68%
60%
42%
40%
22%
14%
20%
0%
Highest HLE Quintile at age 3
© Institute for Fiscal Studies
Read to everyday at age 3
Regular bed times at age 3
Regular meal times at age 3
How much of the socio-economic gap in cognitive
outcomes at age 3 is explained by these factors?
0% 1%
Residual Gap
Parental Education
16%
Family Background/Demographics
1%
34%
3%
Family Interactions
4%
Health and Well-Being
Childcare
Home-Learning Environment
25%
16%
Parenting Style/Rules
Missing Data
Total gap to be explained:
23 percentile points
© Institute for Fiscal Studies
How much of the socio-economic gap in cognitive
outcomes at age 5 is explained by these factors?
Residual Gap
8%
Parental Education
17%
Family
Background/Demographics
52%
Early childhood caring
environment
21%
Missing Data
Prior Ability
0% 1%
Total gap to be explained:
27 percentile points
© Institute for Fiscal Studies
How much of the socio-economic gap in socio-emotional
development at age 3 is explained by these factors?
Total gap to be explained:
22 percentile points
© Institute for Fiscal Studies
How much of the socio-economic gap in socio-emotional
development at age 5 is explained by these factors?
Total gap to be explained:
23 percentile points
© Institute for Fiscal Studies
How much of the socio-economic gap in cognitive
outcomes at age 5 is explained by these factors?
• Gap widens at age (from 23 to 27 percentile points)
• Half of the gap is explained by prior cognitive ability
– Direct effect only: excludes impact via other factors
• 20% via parental education and 17% from family background
• Less than 1% from the early childhood caring environment
• What role for the Home-Learning Environment?
– HLE at age 3 explains age 5 cognitive outcomes through its impact on
age 3 cognitive outcomes
– No impact of age 5 HLE on age 5 cognitive outcomes
• Demonstrates importance of largely pre-determined factors for
outcomes at age 5
© Institute for Fiscal Studies
Summary of early years findings
• Big differences in cognitive development between rich and poor at age
3, widens by age 5
• Children from poor backgrounds face much less advantageous “early
childhood caring environments” than children from better off families.
• Differences in the home learning environment at the age of 3 explain a
substantial proportion of socio-economic gradient
• Larger proportion of the gap remains unexplained, or appears directly
related to other aspects of family background
• Suggests policies to improve parenting skills and home learning
environments in isolation cannot possibly eliminate the cognitive skills
gap between rich and poor young children.
• Wide gaps in socio-emotional development more strongly explained
by differences in early childhood caring environment
© Institute for Fiscal Studies
Primary school years
Paul Gregg and Elizabeth Washbrook (CMPO)
Gaps in educational attainment in the primary
school years
• Average percentile score gap between highest and lowest SEP
quintiles:
– 31 points at 11 (KS2), up from 27 points at 7 (KS1)
– cf. gap of 27 points among MCS children at 5
• Parenting activities and family interactions may continue to
matter, but new potential mechanisms come into play as children
age:
– Parents’ values, beliefs and aspirations for their children
– Children’s own values and beliefs
– Children’s activities and patterns of behaviour
– Experience of schooling
© Institute for Fiscal Studies
Children from poor backgrounds are disadvantaged
across all the mechanisms we consider
• Mother’s locus of control; mother’s valuation of own schooling;
mother’s aspirations for child’s eventual attainment
81% of the highest SEP mothers hope their child will go to
university, compared with 37% of the lowest SEP mothers
• Child’s locus of control; beliefs about own ability; life values
67% of the highest SEP children believe school results are
important in life, compared with 51% of the lowest SEP children
• Anti-social behaviours; hyperactivity and conduct problems;
engagement in leisure activities
• Average Key Stage results and social mix of schools attended
• BUT a few exceptions: mother-child shared activities; child’s
enjoyment of school; teacher-child relations
© Institute for Fiscal Studies
How much of the socio-economic gap at age 11 is
explained by these factors?
-1%
Residual gap
14%
14%
3%
29%
20%
19%
Total gap to be explained:
31 percentile points
© Institute for Fiscal Studies
Parental education and
family background
Child attitudes and
behaviours
Parental attitudes and
behaviours
Pre-school
environments
Schools
Missings
How much of the socio-economic gap at age 11 is
explained by these factors net of prior ability?
Residual Gap
7%
13%
4%
6%
63%
6% 0%
-1%
Parental Education and
Family Background
Child's attitudes and
behaviours
Parent's attitudes and
behaviours
Pre-school environments
Schools
Missing Data
Prior Ability
Total gap to be explained:
31 percentile points
© Institute for Fiscal Studies
Summary of primary years study
• Poor children are much more likely to begin primary school behind
their better-off peers, but even given identical test scores at 7,
poor children fall further behind by age 11
• A wide variety of observable “family process factors” help to
explain the socio-economic gaps at 11 left unaccounted for by
demographic and schooling differences between rich and poor
• Among the multitude of factors identified, some we highlight are
– Mother’s hopes that the child will go to university
– The belief that one’s own actions can make a difference (among both
parents and children)
– Socio-emotional difficulties such as inattention and conduct problems
• Unsurprisingly, positive beliefs and behaviours are strongly related
to test performance at 7 as well as at 11. Nevertheless, the factors
identified appear to contribute to the slower progress of
disadvantaged children even taking their starting point as given.
Secondary school years
Haroon Chowdry, Claire Crawford and Alissa Goodman (IFS)
Outcomes in the secondary school years:
evidence from the LSYPE
• Now focus on educational attainment at 16
• See if gap can be explained by the following characteristics:
– Prior attainment (at 11 and 14)
– Parental education and family background
– School characteristics
– Children’s attitudes and behaviours
• Beliefs and values about being at school
• Aspirations/expectations towards future education
• Behavioural problems
– Parental attitudes and behaviours
• Provision of educational material resources
• Aspirations/expectations about child’s future education
• Home relations and educational interactions with the child
© Institute for Fiscal Studies
Selected differences in characteristics by SEP
Parental expectations by SEP
100%
91%
100%
81%
Per cent
75%
53%
86%
77%
63%
70%
67%
49%
50%
50%
25%
25%
0%
Wants to stay Likely to apply Most friends A job that leads
on in FTE at 16 to HE, and likely likely to stay on somewhere is
to get in
at 16
important
0%
Parent wants YP to stay Parent thinks YP likely
on in FTE at 16
to go to HE
Material resources by SEP
99%
50%
97%
71%
75%
46%
45%
50%
30%
10%
© Institute for Fiscal Studies
21% 24%
6%
1%
5%
8% 10%
8%
8%
0%
0%
Private tuition
41%
20%
10%
25%
Risky behaviours by SEP
40%
Per cent
100%
Per cent
79%
75%
Per cent
76%
Child expectations by SEP
93%
Computer
access
Internet access
Frequent
smoker
Frequent
drinker
Ever tried
cannabis
Ever involved
in anti-social
behaviour
Ever played
truant
How much of the socio-economic gap in cognitive
outcomes at age 16 is explained by these factors?
Residual Gap
7%
6%
15%
59%
8%
1%
4%
Total gap to be explained:
33 percentile points
© Institute for Fiscal Studies
Parental Education and
Family Background
Child's attitudes and
behaviours
Parent's attitudes and
behaviours
Schools
Missing Data
Prior Ability
Summary of secondary years analysis
• Attainment gap at 16 a continuation of earlier gaps
• But might be reduced if poorest children:
– Have access to computer/internet
– Avoid problematic/risky behaviour (in and outside school)
– Expect to go to HE, or have parents who expect them to go
– Believe that they do well in school
• What role for policies to raise education aspirations?
– Aspirations are strongly associated with educational attainment
– Poorest children have lower expectation of going to HE than rich
children, even after taking into account prior attainment
– Suggests an ‘aspirations deficit’ that ought to be alleviated
© Institute for Fiscal Studies
Summary of secondary years analysis
• However:
– HE expectations are already very high across all SEP groups
– Poor children most likely to over-estimate chances of going to HE
90
80
70
60
50
40
30
20
10
0
78%
51%
50%
52%
37%
29%
13%
Poorest
20%
2
Expectations at age 14
© Institute for Fiscal Studies
64%
58%
3
4
Reality at age 18/19
Richest
Children’s cognitive skills: intergenerational
transmission and the socio-economic gap
Claire Crawford, Alissa Goodman and Robert Joyce (IFS)
© Institute for Fiscal Studies
Background and motivation
 Studies of cognitive skills have looked at:
1.
Explanations for the rich-poor socio-economic gap (the rest of this
session!).
2.
Intergenerational transmission (Anger and Heineck, 2009; Bjorklund
et al, 2009; Black et al, 2009).
 Clearly, 1 and 2 could be related. Ideally would like to integrate them
in empirical work.
© Institute for Fiscal Studies
Parental cognitive ability and socioeconomic position
45
% in lowest SEP quintile
40
% in middle SEP quintile
35
% in highest SEP quintile
30
25
20
15
10
5
0
1
3
Parental cognitive ability quintile
© Institute for Fiscal Studies
5
Background and motivation
 Studies of cognitive skills have looked at:
1.
Intergenerational transmission (Anger and Heineck, 2009; Bjorklund
et al, 2009; Black et al, 2009).
2.
Explanations for the rich-poor socio-economic gap (the rest of this
session!).
 Clearly, 1 and 2 could be related. Ideally would like to integrate them
in empirical work.
 If the other papers in this series had observed (e.g.) parental
cognitive ability, would it have...
•
... been an important predictor of cognitive skills, conditional on
other observables?
•
...changed the apparent relative importance of those observables in
explaining the SEP gap?
© Institute for Fiscal Studies
Data
 British Cohort Study (BCS): everyone born in Great Britain
in one week in April 1970 interviewed every few years.
 In age-34 wave, half those who had children were
randomly selected for parent-and-child questionnaires and
children took cognitive tests (BAS).
 So we have:
•Info about the environment children are growing up in.
•Their cognitive test scores.
•Info about the cognitive ability, social skills and attitudes
of their parents when they were children.
© Institute for Fiscal Studies
Measures of parental characteristics
 Cognitive skills
•
BAS scores (word associations, word definitions, pattern
recognition, recall) plus tests of reading, writing, vocab,
maths, copying, sequence recognition at age 10.
•
Also smaller range of similar tests at age 5.
 Noncognitive skills
•
Rutter behaviour scale, ages 5 and 10; Conners behaviour
scale, age 10 (mother-reported).
 Attitudes
•
© Institute for Fiscal Studies
Self-esteem and self-concept measures, ages 10 and 16;
attitudes towards education, age 16 (self-reported).
Sample selection issues
 All children in sample have parent aged 34.
 So children of cohort members who have them after age 34
(31 in our estimation sample) are not included.
 Skews the sample of cohort members (parents)
towards those of lower SEP backgrounds, lower
cognitive ability, lower education levels.
 On other hand, attrition pre-2004 tends to do opposite.
 In terms of observables, these two aspects of nonrandom selection tend to offset each other.
© Institute for Fiscal Studies
Defining the outcome (1)
 We observe BAS scores, as with other papers in series.
 Want to age-standardise them.
 Would typically regress scores on age, and take residuals.
 In our sample, age of child is collinear with age of (a)
parent at child’s birth – and that’s correlated with lots of things
that may affect cognitive test scores.
 Age-standardising in normal way would involve ‘partially’
standardising with respect to SEP, parental ability, etc.
 But we’re interested in the effects of those things!
© Institute for Fiscal Studies
Defining the outcome (2)
 We want to strip out variation in cognitive ability that’s just
due to age. 2 steps:
1. Estimate equation: cogi = agei ’ α + Xi ’ β + ui
2. Define: cogi - agei ’ α
 We take percentile ranks of this.
© Institute for Fiscal Studies
The SEP gap in cognitive test scores: first
decomposition
Young
Person's
Behaviours
10%
Other As
and Bs
8%
Residual gap
16%
CM's gender
0%
Parental
Education
9%
Educational
Attitudes and
Aspirations
23%
Home-learning
Environment
4%
© Institute for Fiscal Studies
Social Skills
14%
Family
Background
16%
The SEP gap in cognitive test scores:
adding new information about parents
Other As and
Bs
8%
Young
Person's
Behaviours
10%
Residual gap
16%
CM's gender
0%
Parental
Education
9%
© Institute for Fiscal Studies
Parental
Cognitive
Ability
16%
Social Skills
11%
Other As and
Bs
4%
Educational
Attitudes and
Aspirations
23%
Homelearning
Environment
4%
Residual gap
6%
CM's gender
Parental
4%
Attitudes and
Parental
Social Skills
Education
8%
5%
Social Skills
14%
Family
Background
16%
Family
Background
12%
Young
Person's
Behaviours
10%
Educational
Attitudes and
Aspirations
20%
Home-learning
Environment
4%
Key findings
 Adding usually unobserved information about parents is
important.
 Predicts about ¼ of SEP gap in cognitive skills.
 Mainly due to parental cognitive ability.
 But reassuringly it does little to change our impression of
relative (predictive) importance of other factors.
 Attitudes and aspirations towards education, family
background, noncognitive skills still important.
© Institute for Fiscal Studies
General conclusions from session (1)
Suggests socio-economic gap in attainment may be reduced
by improving attitudes and behaviours amongst poor children
•Optimistic take would suggest 25% reduction in GCSE
attainment gap
But not a causal analysis. More robust evidence needed to
establish that:
a) attitudes and behaviours can be changed
b) such changes cause improvements in attainment
© Institute for Fiscal Studies
General conclusions from session (2)
Our work suggests that trials may be best focused on:
•
Raising educational aspirations and expectations (for both parents and children) –
and at an earlier stage than e.g. Aim Higher.
•
Supporting the home-learning environment (e.g. pre-school reading).
•
Helping parents and children to believe that their own actions and efforts can help
to improve attainment (locus of control).
Current policy context suggests a disadvantaged ‘pupil premium’ is likely in
the near future.
Might improve educational prospects for the poor, but our work suggests
that focusing on schools in isolation would not eliminate the gap.
© Institute for Fiscal Studies
General conclusions from session (3)
Key message: more evidence needed!
© Institute for Fiscal Studies